{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,8,27]],"date-time":"2026-08-27T05:19:36Z","timestamp":1787807976815,"version":"build-2784847793"},"reference-count":46,"publisher":"IEEE","license":[{"start":{"date-parts":[[2026,6,1]],"date-time":"2026-06-01T00:00:00Z","timestamp":1780272000000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/ieeexplore.ieee.org\/Xplorehelp\/downloads\/license-information\/IEEE.html"},{"start":{"date-parts":[[2026,6,1]],"date-time":"2026-06-01T00:00:00Z","timestamp":1780272000000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-029"},{"start":{"date-parts":[[2026,6,1]],"date-time":"2026-06-01T00:00:00Z","timestamp":1780272000000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-037"}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"published-print":{"date-parts":[[2026,6]]},"DOI":"10.1109\/iwqos70441.2026.11661086","type":"proceedings-article","created":{"date-parts":[[2026,8,26]],"date-time":"2026-08-26T19:09:29Z","timestamp":1787771369000},"page":"1-10","source":"Crossref","is-referenced-by-count":0,"title":["It Takes Two: Embracing Sparsity and Speculative Decoding for Efficient LLM Inference"],"prefix":"10.1109","author":[{"given":"Haolin","family":"Chu","sequence":"first","affiliation":[{"name":"Beijing University of Posts and Telecommunications"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Changyu","family":"Chen","sequence":"additional","affiliation":[{"name":"Beijing University of Posts and Telecommunications"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Wei","family":"Liu","sequence":"additional","affiliation":[{"name":"Independent"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Jian","family":"Luan","sequence":"additional","affiliation":[{"name":"Independent"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Jiabin","family":"Deng","sequence":"additional","affiliation":[{"name":"Beijing University of Posts and Telecommunications"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Liang","family":"Liu","sequence":"additional","affiliation":[{"name":"Beijing University of Posts and Telecommunications"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Huadong","family":"Ma","sequence":"additional","affiliation":[{"name":"Beijing University of Posts and Telecommunications"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Xiaolong","family":"Zheng","sequence":"additional","affiliation":[{"name":"Beijing University of Posts and Telecommunications"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"263","reference":[{"key":"ref1","article-title":"Hydra: Sequentially-dependent draft heads for medusa decoding","author":"Ankner","year":"2024","journal-title":"arXiv preprint arXiv:2402.05109"},{"key":"ref2","article-title":"Speculative streaming: Fast 1lm inference without auxiliary models","author":"Bhendawade","year":"2024","journal-title":"arXiv preprint arXiv:2402.11131"},{"key":"ref3","doi-asserted-by":"publisher","DOI":"10.1609\/aaai.v34i05.6239"},{"key":"ref4","article-title":"Medusa: Simple llm inference acceleration framework with multiple decoding heads","author":"Cai","year":"2024","journal-title":"arXiv preprint arXiv:2401.10774"},{"key":"ref5","article-title":"Accelerating large language model decoding with speculative sampling","author":"Chen","year":"2023","journal-title":"arXiv preprint arXiv:2302.01318"},{"key":"ref6","doi-asserted-by":"publisher","DOI":"10.1109\/TMC.2026.3651722"},{"key":"ref7","doi-asserted-by":"publisher","DOI":"10.1145\/3625687.3625797"},{"key":"ref8","article-title":"Think you have solved question answering? try arc, the ai2 reasoning challenge","author":"Clark","year":"2018","journal-title":"arXiv preprint arXiv:1803.05457"},{"key":"ref9","article-title":"Training verifiers to solve math word problems","author":"Cobbe","year":"2021","journal-title":"arXiv preprint arXiv:2110.14168"},{"key":"ref10","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.acl-long.681"},{"key":"ref11","first-page":"10323","article-title":"Sparsegpt: Massive language models can be accurately pruned in one-shot","volume-title":"In Proceedings of ACM ICML","author":"Frantar"},{"key":"ref12","first-page":"10323","article-title":"Sparsegpt: Massive language models can be accurately pruned in one-shot","volume-title":"In Proceedings of ACM ICML","author":"Frantar"},{"key":"ref13","volume-title":"The language model evaluation harness","author":"Gao","year":"2024"},{"issue":"241","key":"ref14","first-page":"1","article-title":"Sparsity in deep learning: Pruning and growth for efficient inference and training in neural networks","volume":"22","author":"Hoefler","year":"2021","journal-title":"Journal of Machine Learning Research"},{"key":"ref15","article-title":"Determining layerwise sparsity for large language models through a theoretical perspective","author":"Huang","year":"2025","journal-title":"arXiv preprint arXiv:2502.14770"},{"key":"ref16","article-title":"Dynamic low-rank sparse adaptation for large language models","author":"Huang","year":"2025","journal-title":"arXiv preprint arXiv:2502.14816"},{"issue":"S1","key":"ref17","doi-asserted-by":"crossref","DOI":"10.1121\/1.2016299","article-title":"Perplexitya measure of the difficulty of speech recognition tasks","author":"Jelinek","year":"1977","journal-title":"The Journal of the Acoustical Society of America"},{"key":"ref18","first-page":"1725","article-title":"D-1lm: A token adaptive computing resource allocation strategy for large language models","volume-title":"Proceedings of NeurIPS","volume":"37","author":"Jiang"},{"key":"ref19","first-page":"19274","article-title":"Fast inference from transform-ers via speculative decoding","volume-title":"In Proceedings of ACM ICML","author":"Leviathan"},{"key":"ref20","doi-asserted-by":"publisher","DOI":"10.52202\/079017-4487"},{"key":"ref21","doi-asserted-by":"publisher","DOI":"10.52202\/079017-3471"},{"key":"ref22","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.emnlp-main.422"},{"key":"ref23","article-title":"Eagle: Speculative sampling requires rethinking feature uncertainty","author":"Li","year":"2024","journal-title":"arXiv preprint arXiv:2401.15077"},{"key":"ref24","article-title":"Bita: Bi-directional tuning for lossless acceleration in large language models","author":"Lin","year":"2024","journal-title":"arXiv preprint arXiv:2401.12522"},{"key":"ref25","article-title":"Parallel speculative decoding with adaptive draft length","author":"Liu","year":"2024","journal-title":"arXiv preprint arXiv:2408.11850"},{"key":"ref26","first-page":"22137","article-title":"Deja vu: Contextual sparsity for efficient 1lms at inference time","volume-title":"In Proceedings of ICML","author":"Liu"},{"key":"ref27","year":"2023","journal-title":"Vicuna v1.5 model"},{"key":"ref28","volume-title":"Evaluating large language models trained on code","author":"Mark Chen","year":"2021"},{"key":"ref29","volume-title":"Llama2-7b-chat-hf","year":"2023"},{"key":"ref30","volume-title":"Pytorch cuda memory allocated api","author":"Paszke","year":"2025"},{"key":"ref31","volume-title":"Big o notation","author":"Paul Bachmann","year":"1894"},{"key":"ref32","doi-asserted-by":"publisher","DOI":"10.1145\/3474381"},{"key":"ref33","volume-title":"Statproofbook\/statproof-book.github.io: Statproofbook 2023","author":"Soch","year":"2023"},{"key":"ref34","doi-asserted-by":"publisher","DOI":"10.1145\/3694715.3695964"},{"key":"ref35","article-title":"A simple and effective pruning approach for large language models","author":"Sun","year":"2023","journal-title":"arXiv preprint arXiv:2306.11695"},{"issue":"6","key":"ref36","first-page":"7","volume-title":"Alpaca: A strong, replicable instruction-following model. Stanford Center for Research on Foundation Models","volume":"3","author":"Taori","year":"2023"},{"key":"ref37","volume-title":"Stanford alpaca: An instruction-following llama model","author":"Taori","year":"2023"},{"key":"ref38","year":"2024","journal-title":"Qwen2.5: A party of foundation models"},{"key":"ref39","article-title":"Quantspec: Selfspeculative decoding with hierarchical quantized kv cache","author":"Tiwari","year":"2025","journal-title":"arXiv preprint arXiv:2502.10424"},{"key":"ref40","article-title":"Q-sparse: All large language models can be fully sparsely-activated","author":"Wang","year":"2024","journal-title":"arXiv preprint arXiv:2407.10969"},{"key":"ref41","doi-asserted-by":"publisher","DOI":"10.1145\/3695053.3730996"},{"key":"ref42","article-title":"Powerinfer-2: Fast large language model inference on a smartphone","author":"Xue","year":"2024","journal-title":"arXiv preprint arXiv:2406.06282"},{"key":"ref43","article-title":"Outlier weighed layerwise sparsity (owl): A missing secret sauce for pruning llms to high sparsity","author":"Yin","year":"2023","journal-title":"arXiv preprint arXiv:2310.05175"},{"key":"ref44","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/P19-1472"},{"key":"ref45","doi-asserted-by":"publisher","DOI":"10.52202\/075280-2020"},{"key":"ref46","doi-asserted-by":"publisher","DOI":"10.52202\/079017-0758"}],"event":{"name":"2026 IEEE\/ACM International Symposium on Quality of Service (IWQoS)","location":"Istanbul, Turkiye","start":{"date-parts":[[2026,6,29]]},"end":{"date-parts":[[2026,7,2]]}},"container-title":["2026 IEEE\/ACM International Symposium on Quality of Service (IWQoS)"],"original-title":[],"link":[{"URL":"http:\/\/xplorestaging.ieee.org\/ielx8\/11660881\/11661011\/11661086.pdf?arnumber=11661086","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,8,27]],"date-time":"2026-08-27T04:41:40Z","timestamp":1787805700000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/11661086\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,6]]},"references-count":46,"URL":"https:\/\/doi.org\/10.1109\/iwqos70441.2026.11661086","relation":{},"subject":[],"published":{"date-parts":[[2026,6]]}}}